A 10-Gbps, 0.121-pJ/bit, All-Digital True Random-Number Generator using Middle Square Method

The rapid development of mobile and cloud markets has resulted in an exponential increase in sensitive data such as biometric or financial information. Therefore, the need for random-number generators with high entropy for encryption is increasing. For mass-production at a low cost and easy integration into a general system-on-chip (SoC), true random-number generators (TRNGs) are often fabricated in commercial CMOS processes. Most CMOS-based TRNGs use either a jitter- or metastability-based entropy source (ES) and combine it with a post-processor to improve the random-number quality. Although a TRNG using a jitter-based ES [1] has a small area and low power consumption, it is weak against power supply attack (PSA). Therefore, additional circuits are required to prevent such an attack. Moreover, it is slow in terms of speed. Meanwhile, the use of a metastability-based ES increases the power consumption of a TRNG; however, it can generate random numbers at a faster rate and is robust against PSA. In this work, we propose a high-throughput TRNG using the middle square method (MSM). To lower the post-processing power, a robust digital comparator with offset calibration is used as an ES, which allows simplification of the MSM post-processing.